Systems and methods for predicting glycosylation on proteins

Abstract

The disclosed technology provides a computational prediction modeling comprising a novel algorithm for prediction of glycosylation or to optimize biopharmaceutical production of proteins of therapeutic relevance. The model of the disclosed technology can be used to predict glycosylation changes based solely on the stating glycoprofiles in any host cells and known or suggested rules on enzyme specificity. Applications of the invention model are also provided.

What is claimed is: 
     
       1. A method comprising:
 generating, by a computing device, using a known initial structure and a plurality of known glycosylation enzymes and/or reactions, a generic glycosylation reaction network comprising a plurality of glycans; 
 receiving, at the computing device, information relating to a measured glycoprofile of a particular protein, the measured glycoprofile comprising a plurality of glycoprofile glycans, each having an associated relative frequency; 
 tailoring the generic glycosylation reaction network to the measured glycoprofile to provide a tailored network, the tailored network comprising a reduced set of the plurality of glycans, the reduced set including the glycoprofile glycans; 
 transforming the tailored network into a Markov chain, the Markov chain representing a stochastic network wherein each glycoprofile glycan in the reduced set is regarded as a state in the stochastic network that can transition to another state in the stochastic network with a particular transition probability; 
 determining reaction conditions that approximate the particular transition probabilities that define a set of desired states in the stochastic network; and 
 culturing cells under the reaction conditions that approximate said particular transition probabilities, thereby producing a desired glycoprofile glycan. 
 
     
     
       2. The method of  claim 1 , wherein the plurality of glycans can be synthesized (i) by combined action of the plurality of known enzymes or (ii) through chemical reactions. 
     
     
       3. The method of  claim 1 , wherein the known initial structure is Man 9 GlcNAc 2 . 
     
     
       4. The method of  claim 1 , wherein the generic glycosylation reaction network is a flux-balance network; and the associated relative frequencies are obtained experimentally and wherein the associated relative frequencies are obtained experimentally through at least one of mass spectrometry-coupled liquid chromatography (LC-MS), liquid chromatography, and liquid chromatography lectin arrays. 
     
     
       5. The method of  claim 1 , wherein tailoring the generic glycosylation reaction network comprises identifying a plurality of reactions not required to obtain the plurality of glycoprofile glycans. 
     
     
       6. The method of  claim 1 , wherein the tailored network represents a reaction topology describing how flux from a known initial structure leads to generation of the measured glycoprofile, and wherein the tailored network provides a minimal reaction network for producing the measured glycoprofile. 
     
     
       7. The method of  claim 1 , wherein tailoring the generic glycosylation reaction network comprises using model reduction through convex analysis or optimization algorithms. 
     
     
       8. The method of  claim 1 , wherein transforming the tailored network into a Markov chain comprises:
 assessing variance in reaction fluxes of the tailored network using Monte Carlo sampling, the Monte Carlo sampling producing a plurality of flux vectors having an associated variance, wherein each reaction flux represents a path through the tailored network from the known initial structure to the measured glycoprofile; and 
 reformulating each of the plurality of flux vectors into a plurality of associated transition probabilities. 
 
     
     
       9. The method of  claim 1 , further comprising:
 simulating a perturbation in the tailored network by modifying the Markov chain to determine how the perturbation will affect the measured glycoprofile. 
 
     
     
       10. The method of  claim 9 , wherein the perturbation is a decrease in activity of a particular enzyme and simulating the decrease in the activity of the particular enzyme comprises:
 identifying a first subset of reactions in the tailored network that depend on the particular enzyme, a second subset of reactions in the tailored network representing the remaining reactions in the tailored network that do not depend on the particular enzyme; 
 scaling down, by a user-provided factor, the first subset of reactions to provide a set of scaled-down reactions; 
 modifying, based on the scaled-down reactions, transition probabilities associated with the second subset of reactions; and 
 generating an updated Markov chain yielding a predicted glycoprofile gained through the decrease in activity of the particular enzyme. 
 
     
     
       11. The method of  claim 9 , wherein the perturbation is an upregulation in activity of a particular enzyme and simulating the upregulation in the activity of the particular enzyme comprises:
 identifying a first subset of reactions in the tailored network that depend on the particular enzyme, a second subset of reactions in the tailored network representing the remaining reactions in the tailored network that do not depend on the particular enzyme; 
 scaling up, by a user-provided factor, the first subset of reactions to provide a set of scaled-up reactions; 
 modifying, based on the scaled-up reactions, transition probabilities associated with the second subset of reactions; and 
 generating an updated Markov chain yielding a predicted glycoprofile gained through the upregulation in activity of the particular enzyme. 
 
     
     
       12. The method of  claim 1 , wherein the Markov chain has an absorption probability, and the absorption probability provides the Markov chain's probability of reaching an absorbing state from the known initial structure. 
     
     
       13. The method of  claim 1 , wherein the generic glycosylation reaction network describes at least one selected from the group comprising N-glycosylation, O-glycosylation, oligosaccharide synthesis, synthesis of glucosaminoglycans, glycolipids, proteoglycans, glycoconjugates, and peptidoglycans. 
     
     
       14. The method of  claim 1  further comprising determining an optimal combination of changes to match a desired glycoprofile by:
 providing an optimization algorithm for implementing an optimization process that simulates possible perturbations that lead to a desired glycoprofile, the optimization algorithm minimizing an objective function that measures a distance between a predicted and an observed glycoprofile; 
 receiving an initial hypothesis of perturbations and a stopping condition; and 
 determining an end point of the optimization process, wherein the end point represents the optimal combination of changes and comprises a set of perturbations that provides a simulated glycoprofile within a specified margin of error of the desired glycoprofile. 
 
     
     
       15. The method of  claim 14 , wherein the simulated possible perturbations reduce or increase enzyme activity, change enzyme substrates or change media conditions. 
     
     
       16. The method of  claim 1 , further comprising:
 a) tailoring the generic glycosylation reaction network for alternative localization; 
 b) separately simulating alternatives and comparing resulting predicted glycoprofiles to a desired glycoprofile; and 
 c) changing localization of glycosylation enzymes based at least in part on the separately simulated alternatives and comparison of resulting predicted glycoprofiles to the desired glycoprofile of step b). 
 
     
     
       17. The method of  claim 16 , wherein predictions are used to change localization of glycosylation enzymes to change the glycoprofile. 
     
     
       18. One or more computer-readable media having stored thereon executable instructions that when executed by one or more processors configure a computer system to perform the method according to  claim 1 . 
     
     
       19. A system comprising:
 a computing device; 
 a memory device; and 
 a cell culturing device, 
 wherein the memory device has stored thereon executable instructions that when executed by one or more processors of the computing device configure the computing device to perform the method according to  claim 1 , and 
 wherein the cell culturing device is configured to culture cells under the reaction conditions that approximate the particular transition probabilities, thereby producing the desired glycoprofile glycan corresponding to the state in the stochastic network based on instructions from the computing device. 
 
     
     
       20. The method of  claim 1 , wherein the reaction conditions under which the culturing of the cells is performed include a genetic modification of the cells. 
     
     
       21. The method of  claim 1 , wherein the reaction conditions under which the culturing of the cells is performed include a modification of a culture medium in which the cells are cultured.